Prediction of birth weight using fetal thigh and upper-arm volumes by three-dimensional ultrasonography in a Brazilian population

Abstract

To assess the accuracy of three-dimensional (3DUS) upper-arm (VolArm) and thigh (VolTh) volume measurements in the prediction of birth weight (BW). A cross-sectional study involving 81 live singletons was performed. VolArm and VolTh were obtained using 3DUS multiplanar mode with 5 mm slices. Linear and polynomial regressions were calculated to determine the best formula to predict BW using VolArm and VolTh. Analysis of variance was used to compare errors in BW using these formulae and using Shepard's and Hadlock's formulae. The interclass correlation coefficient (ICC) was used to assess intra and interobserver variability of measurements. The best formula to predict BW based on VolArm was a simple linear equation (BW = 803.91 + 39.89VolArm), and for VolTh it was a second degree polynomial equation (BW = 32.37VolTh - 0.06VolTh(2)). A third formula using both parameters was also constructed (BW = 792.87 + 22.81VolArm + 7.54VolTh). The error (E), percent error (PE), absolute error (AE) and absolute percent error (APE) for the BW prediction using VolArm were 0 g, 163.4 g, 0.5% and 5.4%, respectively. The same results for E, PE, AE and APE using VolTh were 0.99 g, 155.5 g, 0.3% and 5.2% and for the combined formula using both VolArm and VolTh the E, PE, AE and APE were 0 g, 138.4 g, -0.4% and 4.6%, respectively. VolArm and VolTh were highly reproducible with intraobserver ICC of 0.98 and 0.99 and interobserver ICC of 0.96 and 0.97, respectively. BW estimated through formulae that use the fetal arm and thigh volumes assessed through 3DUS are not superior to two-dimensional formulae.